Get a small number of UMI per cell

[dongspy]

Using the h5 file(GSM5151369_PD212_scifi_2C_4lines_7650_MeOH-cells.h5ad.gz) uploaded to the GEO(https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSM5151369), we analyzed the median genes per cell per plate_well.
Note that, in each plate_well, the top 500 cells with the largest number of UMI were used.

But we get a small number of UMI per cell that are not consistent with the figure 2e. The maximum median UMI count per plate_well is 89.

The table below is the median UMI count per plate_well.

scifi_h5_UMIcount_plate.csv

Below is my code.

import numpy as np
import pandas as pd
import scanpy as sc

h5_file = "GSM5151369_PD212_scifi_2C_4lines_7650_MeOH-cells.h5ad"

adata = sc.read_h5ad(h5_file)
sc.pp.calculate_qc_metrics(adata, percent_top=None, log1p=False, inplace=True)
# top_cells = 500

plate_well_gene_counts_df = adata.obs.groupby('plate_well').apply(
    lambda df: df['total_counts'].sort_values(ascending=False).head(500).median()).reset_index()

plate_well_gene_counts_df.columns = ['plate_well', 'Median UMI Count']
plate_well_gene_counts_sort_df = plate_well_gene_counts_df.sort_values('Median UMI Count', ascending=False)
plate_well_gene_counts_sort_df.to_csv('data/scifi_h5_UMIcount_plate.csv', index=False)

Is there a problem with this analysis ？